Unified machine-learning-based design method for normal and high strength steel I-section beam–columns

• DOI: 10.1016/j.tws.2024.111835
• Publication Date: 2024-03-27T03:49:44Z
• Authors: Andi Su; Jinpeng Cheng; Xuelai Li; Yukai Zhong; Shuai Li; Ou Zhao; Ke Jiang
• Keywords: Hyperparameter; Boosting; Gradient boosting

High strength steel is regarded as a promising construction material due to its superior mechanical properties. However, the codified failure load predictions for high welded I-section beam–columns are not accurate in some cases owing lack of relevant design codes structures. In addition, current ignore interaction effect cross-section plate elements, leading inaccuracy predictive ultimate resistances both normal and beam–columns. this paper, unified method was proposed based on machine learning failing by different modes. Firstly, total 812 experimental numerical data with various grades, geometric dimensions, including dimensions member lengths, modes, global buckling, local buckling local–global interactive were collected establish database. Seven algorithms, Decision Tree, Random Forest, Support Vector Machine, K-Nearest Neighbour, Adaptive Boosting, Extreme Gradient Boosting Categorical applied develop regression models predict loads. Based established database, each model then trained key hyperparameters optimised. The performance evaluated through series statistical indicators feature importance analysis, results indicated that XG-trained had highest level accuracy. Finally, given compared those existing provisions, Eurocode American codes. evaluation bearing capacities scattered inaccurate, while provided substantially improved predictions.